Abstract: This paper aims at the problem of tracking user’s interest drift and optimizing user profile. It proposes a method of drifting user’s interests based on time window optimization. User profile tracks user’s interest drift by classification error-rate and when user’s interests are changed, user profile optimizes time window size automatically. Experiment shows that this method can describe the interest change of the user exactly and improve the personalization.

Key words: interest drift, classification error-rate, time window optimization

摘要： 针对个性化服务系统中如何跟踪用户兴趣变化和调整用户兴趣模型的问题，提出基于优化时间窗的用户兴趣漂移算法。利用分类错误率的变化跟踪用户兴趣的漂移，当用户兴趣发生变化时，通过优化时间窗算法自动调节时间窗的大小，以达到调整用户兴趣模型的目的。实验表明，该方法能较为准确地描述用户的兴趣变化，提高个性化信息服务的效率。

关键词: 兴趣漂移, 分类错误率, 优化时间窗

CLC Number: 

• TP391